Systematic analysis reveals molecular characteristics of erg-negative prostate cancer

ABSTRACT The _TMPRSS2:ERG_ gene fusion is the most prevalent early driver gene activation in prostate cancers of European ancestry, while the fusion frequency is much lower in Africans and

Asians. The genomic characteristics and mechanisms for patients lacking _ERG_ fusion are still unclear. In this study, we systematically compared the characteristics of gene fusions, somatic

mutations, copy number alterations and gene expression signatures between 201 _ERG_ fusion positive and 296 ERG fusion negative prostate cancer samples. Both common and group-specific

genomic alterations were observed, suggesting shared and different mechanisms of carcinogenesis in prostate cancer samples with or without _ERG_ fusion. The genomic alteration patterns

detected in _ERG_-negative group showed similarities with 77.5% of tumor samples of African American patients. These results emphasize that genomic and gene expression features of the

_ERG_-negative group may provide a reference for populations with lower _ERG_ fusion frequency. While the overall expression patterns were comparable between _ERG_-negative and

_ERG_-positive tumors, we found that genomic alterations could affect the same pathway through distinct genes in the same pathway in both groups of tumor types. Altogether, the genomic and

molecular characteristics revealed in our study may provide new opportunities for molecular stratification of _ERG_-negative prostate cancers. SIMILAR CONTENT BEING VIEWED BY OTHERS

EARLY-ONSET METASTATIC AND CLINICALLY ADVANCED PROSTATE CANCER IS A DISTINCT CLINICAL AND MOLECULAR ENTITY CHARACTERIZED BY INCREASED _TMPRSS2–ERG_ FUSIONS Article 08 January 2021 PROSTATE

CANCER SUBTYPING AND DIFFERENTIAL METHYLATION ANALYSIS BASED ON THE ETS FAMILY OF TRANSCRIPTION FACTORS FUSION GENES Article Open access 06 September 2024 GENETIC ALTERATIONS IN THE

3Q26.31-32 LOCUS CONFER AN AGGRESSIVE PROSTATE CANCER PHENOTYPE Article Open access 14 August 2020 INTRODUCTION Prostate cancer is the second most commonly diagnosed cancer type in men

globally and the fifth leading cause of cancer death, accounting for 6.6% of death among men1. Significant efforts have been made to characterize recurrent genomic alterations in prostate

cancers, which may be potential driver events2,3,4,5. The overall mutation burden in prostate cancer is relatively low (0.3–2 non-synonymous somatic mutations per megabase) compared to other

cancer types2,6,7. The most common genomic alteration is the fusion of 5′-UTR of _TMPRSS2_ (21q22) with 3′-end of ETS family members, such as _ERG_ (21q22), _ETV1_ (7p21), _ETV4_ (17q21),

or _ETV5_ (3q27)8,9,10,11. Significantly mutated genes include _SPOP_, _FOXA1_, _TP53_, _MED12_, and _CDKN1B_2,5,12. In addition to somatic mutations, somatic copy number alterations (SCNA)

are recurrently seen in prostate cancer, including the amplification of chromosome 7 and 8q (affecting the _MYC_ locus), and the focal deletion of chromosome 1q42, 3p13 (_FOXP1_), 4p15,

6q12–22 (_MAP3K7_), 8p, 13q, 16q, 17p (_TP53_), 18q12, and 21q22.3 (_TMPRSS2-ERG_ fusion)5,7,12,13. However, there is still a large proportion of prostate cancer genomes that remains to be

evaluated5,14,15. Further studies confirmed that the _TMPRSS2_-_ERG_ fusion is caused by an interstitial deletion on chromosome 21 or by a chromosomal translocation. These genomic

rearrangements results in the overexpression of the _ERG_ oncogene and ERG oncoprotein16,17. A variety of biological processes and pathways including cell invasion, Androgen receptor (AR)

signaling, Transforming growth factor beta 1 (TGF-β) signaling have been implicated in _ERG_ dysregulation18,19,20,21,22. _ERG_ oncogenic activation is an early causal event in prostate

cancer23,24,25. In some reports _TMPRSS2-ERG_ fusion is positively correlated with advanced tumor stage, high Gleason score, and worse survival17,26,27,28,29,30. While some studies did not

found significant association between _ERG_ fusion and disease progression26,31,32,33,34, numerous studies reported positive correlation of ERG-negative prostate tumor type with disease

progression35,36,37. Since _TMPRSS2_-_ERG_ fusion is a dominant molecular subtype in prostate cancer in European descents, it provides opportunities for targeted cancer therapy. Along these

lines, direct and indirect _ERG_ targeted therapeutic approaches are being developed38,39,40. Patients harboring ERG oncoprotein positive tumors are more likely to benefit from ERG targeted

therapy. However, the frequency of _TMPRSS2-ERG_ fusion significantly varies in different ethnic groups41. African American (20%~30%) and Asian (less than 20%) has much lower fusion

frequency compared to Caucasian (~50%)42,43,44. In contrast to _ERG_ fusion positive tumors, the genomic characteristics are not yet clear for the _ERG_ fusion negative tumor type45.

Therefore, identification of driver events in _ERG_-negative prostate cancer is important for understanding the mechanism of tumorigenesis. In this study, we systematically explored the

genomic and molecular differences of gene fusions, somatic mutations, SCNAs, gene expression signatures and dysregulation of pathways in prostate tumors with or without _ERG_ fusion using

publicly available data. Our results provide new insights into the molecular landscape highlighting specific mechanisms of prostate tumorigenesis. RESULTS DATA SOURCES AND THE RELATIONSHIP

BETWEEN _ERG_ FUSION, DELETION, AND EXPRESSION We collected the _ERG_ fusion status information from two prostate cancer genome studies and compared the relationship among _ERG_ fusion,

deletion and expression5,46. The two datasets were highly consistent, except for 13 samples where the fusion status was unclear in the genomic data (Fig. 1a). We checked the _ERG_ expression

in these 13 samples and found significantly higher expression compared with _ERG_ fusion negative samples (t.test, p-value = 0.001), indicating _ERG_ may be activated in these 13 samples.

Therefore, we assigned a sample into the ERG-positive group if its ERG fusion was detected in either study. As a result, we identified 201 _ERG_-positive samples and 296 _ERG_-negative

samples for subsequent analysis (Supplementary Table 1). Also, we used _ERG_ gene expression to verify the genomic classification of samples. Since, _ERG_ fusion could result from either

translocation or deletion at 21q22.316, we found that 40.8% _ERG_-positive samples harbored _ERG_ deletion. Clinical characteristics of the _ERG_-positive and _ERG_-negative groups are

summarized in Table 1. Although, patients with higher Gleason score (4 + 3 or 8–10) were more frequently found in the _ERG_-negative group, biochemical recurrence-free survival of patients

showed no difference between the two groups (Supplementary Fig. 1, p-value = 0.29, Log-rank test). The TCGA prostate cancer cohort contained 279 Caucasian American (CA), 40 African American

(AA), 5 Asian men and 173 without known ancestry. The proportion of ERG-positive samples in CA was higher than that in AA (47% vs. 35%, Fig. 1b), which is in accordance with previous studies

(Supplementary Table 2). Like TCGA, most of the previous studies focused on patients of European ancestry. Indeed, more studies are needed for African and Asian patients that harbor mostly

ERG-negative prostate cancers. COMMON AND SPECIFIC GENOMIC ALTERATIONS IN _ERG_-POSITIVE AND _ERG_-NEGATIVE PROSTATE CANCERS GENE FUSIONS Consistent with previous studies47,48, in

_ERG_-positive group, the most frequent fusion partner of _ERG_ in our study was _TMPRSS2_ (94.1%), and the second was the _SLC45A3_ gene (6.4%, located at 1q32.1, Fig. 2a). These two genes

both have AR responsive promoter and share similar mechanisms in _ERG_ overexpression48. As expected, significantly higher _ERG_ expression was detected in samples harboring _SLC45A3:ERG_

fusion compared with samples with non-detectable _ERG_ fusion (pvalue = 5e-5, one-tailed t.test). Other two ETS-family members, _ETV1_ and _ETV4_, show relatively high genomic rearrangement

frequencies in _ERG_-negative group (4.7% and 2.7%, respectively). We found that the _LSAMP_ gene that is frequently deleted in ERG-negative prostate tumors of African American men49, was

often rearranged including fusion with _ZBTB20_ specifically in the _ERG_-negative group. Moreover, tumor suppressor gene _MIPOL1_ and _TTC6_ fusion were also specifically detected in the

_ERG_-negative group at notable frequency (3.7%, Fig. 2a). Recent study of 65 Chinese prostate cancer whole genomes also reported _TTC6:MIPOL1_ fusion detected at 6.2% frequency44. Indeed,

detection of _TTC6:MIPOL1_ fusion may have potential implication for prostate cancers of non-European ancestry. In addition, ten of eleven recurrent gene fusions (detected at least in three

samples) have been reported in other literatures. Thus prostate cancer genomic fusions detected in our study, as well as in other reports are more likely real than false positives

(Supplementary Table 3). SOMATIC MUTATIONS We used MutSigCV to identify significantly mutated genes in the _ERG_-positive and _ERG_-negative groups respectively50. Only two genes, _TP53_ and

_PTEN_, were significantly mutated in _ERG_-positive group. By contrast, eight genes were significantly mutated in the _ERG_-negative group (Fig. 2b). In addition to known recurrently

mutated genes _SPOP_ and _FOXA1_ which were reported to be mutually exclusive with _ERG_ rearrangements2,5, we found that the mutation frequency of _CDK12_ and _KDM6A_ were significantly

higher in the _ERG_-negative group (Fig. 2b, p-value = 1.18e-3 and 3.26e-4, respectively. Fisher.test). SOMATIC COPY NUMBER ALTERATIONS We applied the GISTIC algorithm to discern significant

copy number alterations in the _ERG_-positive and _ERG_-negative groups51. First, we assessed the overall distribution of copy number alterations of all prostate cancer genomes in our study

(Fig. 2c,d). Overall, deletions were more commonly than amplifications showing similar distribution in both _ERG_-positive and _ERG_-negative groups. Copy number alterations affected

similar regions within the two groups, while deletion and amplification frequencies showed variations. Twenty one amplified regions including chromosome 8q, 11q13, 14q21, 16q11, 1q22, 3q26

and 17q23, were recurrently altered in the _ERG_-negative group (Supplementary Table 4, residual q value < 0.05). The _ERG_-positive group harbored similar amplified regions, but did not

reach statistical significance due to lower frequencies. Among the regions of copy number gains, chromosome 8q that includes the _MYC_ oncogene exhibited a relatively high frequency (~40%).

In another complex CNV region at 14q21.1 spanning _MIPOL1/FOXA1/TTC6_ locus, the _MIPOL1:TTC6_ gene fusion was detected. Moreover, we found several chromosome arm-level amplifications with

significantly higher frequency in the _ERG_-negative tumors than in _ERG_-positives, including chromosome 8 (38.5% vs. 19%) and chromosome 7 (26.1% vs. 11.5%) (Fig. 2c). Ten regions were

commonly deleted in both _ERG_-positive and _ERG_-negative groups, including 6q14.3, 13q14.13, 10q23.31, 12p13.1, 5q11.2, 5q13.2, 17p13.1, and 16q22.3 (residual q < 0.05), which is

consistent with previous reports5,13. Twenty two and twenty five copy number losses were detected only in the _ERG_-positive or in the _ERG_-negative group, respectively. Among these focal

deleted regions, some showed significantly different frequency between the two groups. Similar to previous studies we also detected frequent deletions of 21q22 (_ERG_, _TMPRSS2_), 17p13.1

(_TP53_), and 10q23.31 (_PTEN_) in _ERG_-positive tumors, while 6q14.3 and 13q14.13 deletions were more frequent in _ERG_-negatives (Fig. 2d). Additionally, two novel regions, 6q16.3

(_HACE1_) and 6q22 (_FRK_) were deleted more frequently in the _ERG_-negative group. To gain more insight into the functional effects of SCNA regions, we assessed the genomic defects of

tumor suppressor genes (TSGs) and oncogenes (Supplementary Table 5). Thirty-two TSGs were recurrently altered with frequencies higher than 20% in both groups (Fig. 2e). Twenty-one (65.6%) of

these genes were previously shown to play roles in the progression of prostate cancer. Other genes with high alteration frequencies need to be further defined. Thirteen TSGs and one

oncogene showed significantly higher alteration frequency in the _ERG_-negative group, and another thirteen tumor suppressor genes and one oncogene showed significantly higher alteration

frequency in ERG-positives (Fig. 2f). The candidate CNV genes found in TCGA dataset show comparable alteration frequency in an independent whole genome sequencing dataset, which includes 7

_ERG_-positive and 7 _ERG_-negative prostate tumors (Supplementary table 5, CPDR dataset). Among these group-specific SCNA genes, we found that ten genes were significantly associated with

biochemical recurrence. In addition to previously reported disease progression related genes _TP53_, _PTEN_ and _FOXP1_, we also found that an additional seven genes were associated with

biochemical recurrence (Supplementary Fig. 2). Although _ERG_ rearrangement status alone might not be a definitive marker for disease progression, our findings highlight a subset of genes

associated with higher risk of disease progression (Overall prevalence: 44.87%). Furthermore, we found a group of tumor suppressor genes including _FRK_, _WISP3_, _PRDM1_, and _LRP1B_ whose

CNV and expression may indicate interactions with known drugs and therefore, are potentially actionable (Supplementary Fig. 3). CANDIDATE GENES ASSOCIATE WITH GENOMIC ALTERATION PATTERNS IN

ERG-NEGATIVE PROSTATE TUMORS Since we have characterized both common and group-specific genomic alterations with high frequency in _ERG_-positive and _ERG_-negative prostate tumors, we next

examined the molecular portrait of the _ERG_-negative group based on the associated candidate genes. First, we combined the genomic alterations of gene fusions, somatic mutations and copy

number alterations which occur recurrently in the _ERG_-negative group. Next, we removed the redundant alterations to find a subset of genes highly represented in the genomic alteration

pattern of _ERG_-negative tumors. Nine representative genes have emerged from the analysis (Fig. 3a). Genomic alteration of one or more of these nine genes were detected in 67.7% of the

_ERG_-negative group. Since _ERG_-rearrangement are less frequent in prostate cancers of African descents, we explored whether candidate gene defects found in the _ERG_-negative group are

present or absent in prostate cancers of AA men. As _ERG_ is less frequent in prostate cancers of AA patients, we evaluated the alteration patterns of the nine genes characteristic to

_ERG_-negative tumors in available datasets of 40 AA prostate tumor samples. We found that 77.5% AA tumors harbor at least one of the nine gene signatures associated with _ERG_-negative

tumors indicating similar patterns between prostate cancers of AA patients and the overall genomic alteration pattern of ERG-negative tumors (Fig. 3b). Among the nine representative genes,

_NKX3-1_, _RB1_, and _CDH13_ were commonly deleted in both _ERG_-positive and _ERG_-negative tumors. Other genes had significantly more alterations in _ERG_-negative samples. The oncogene

_MYC_ mRNA is up-regulated in tumor compared to normal. Tumors with _MYC_ amplification show significantly higher expression of _MYC_ gene and higher probability of disease progression than

other patients (Fig. 3c,d). The Zinc finger transcription factor, _ZNF292_ was shown to function as a tumor suppressor in gastric cancer, colorectal cancer, and chronic lymphocytic

leukemia52,53. Deletion of ZNF292 in prostate cancer results in decreased expression (Fig. 3e), which may promote tumor development. COMPARISON OF METHYLATION AND EXPRESSION BETWEEN

_ERG_-POSITIVE AND _ERG_-NEGATIVE TUMORS Since promoter hypermethylation is widely observed in multiple cancers, we investigated the hypermethylated sites in promoter regions (TSS200,

TSS1500, 5’UTR and 1stExon) of genes with low mRNA expression (See Method). Compared to normal samples, 2191 CpG sites (694 genes) and 1871 CpG sites (645 genes) were hyper-methylated in

_ERG_-positive and _ERG_-negative groups, respectively. Approximately 70% of them were overlapped between the two groups (Supplementary Fig. 4a). Direct comparison between two tumor groups

indicated 51 hyper-methylated sites (31 genes) in _ERG_-negative and 14 hyper-methylated sites (8 genes) in _ERG_-positive tumors (Supplementary Fig. 4b). Therefore, the overall methylation

profiles showed similarities between the two groups. We compared the expression profiles of _ERG_-positive, _ERG_-negative tumor and prostate tissue samples with morphologically normal

appearance to identify differentially expressed genes among these three groups. A large proportion (>70%) of differentially expressed (DE) genes were common in the _ERG_-positive and

negative groups (Fig. 4a). As expected, common DEs were significantly enriched in essential pathways like calcium signaling and cAMP signaling pathways (Fig. 4b). Common up-regulated genes

were significantly enriched in cell cycle which is recurrently altered in cancer. However, no significant functional GO term was enriched for group-specific genes indicating comparable

expression profiles between _ERG_-positive and _ERG_-negative prostate tumor types, despite in their differences in their dominant driver genomic alterations. These findings indicate that

different genomic alternations may have similar effects on gene expression, resulting in similar phenotype. THE IMPACT OF GENOMIC ALTERATIONS ON PATHWAY DYSREGULATION IN _ERG_-POSITIVE AND

_ERG_-NEGATIVE PROSTATE TUMORS We selected eleven pathways either cancer-related or reported to be important in prostate cancer from Misgdb54,55,56. Next, we compared the frequency of CNV,

somatic mutation and gene fusion of the _ERG_-positive and _ERG_-negative groups based on publicly available TCGA data. The male hormone axis (AR pathway) was the only node that altered

significantly more frequently in _ERG_-positive group that is consistent with the AR regulation of _ERG_ in the context of _TMPRSS2:ERG_ fusion (Fig. 5a). However, there were still 65.3% of

_ERG_-negative samples with AR pathway disruption, which were apparently affected by other genes in the AR pathway (Fig. 5b). For example, _CDK6_ (10.7% _ERG_-negative vs. 4% ERG-positive),

_NCOA2_ (23.7% _ERG_-negative vs. 10.5% _ERG_-positive) and _PRKDC_ (20.0% _ERG_-negative vs. 11.5% _ERG_-positive). Similarly, some component of NOTCH signaling pathway signatures had

higher alteration frequency in the _ERG_-positive group (e.g., _DVL2_,11.3% _ERG_-negative vs. 25.0% _ERG_-positive) while _HDAC2_ (24.4% _ERG_-negative vs. 7.0% _ERG_-positive) had higher

alteration frequency in _ERG_-negative group (Fig. 5c). They both inhibit NOTCH signaling pathway but function at different contexts. Therefore, the observed prostate cancer genomic and

expression alterations of different genes may affect the same pathway resulting in comparable expression profiles between _ERG_-positive and _ERG_-negative prostate tumor types. DISCUSSION

Our study provides new insights into the molecular landscape of _ERG_-negative prostate cancers. Except for known alterations mutually exclusive with _ERG_ rearrangements, such as mutation

in _SPOP_ and _FOXA1_, we found that gene fusion of _TTC6:MIPOL1_ and somatic mutation on _CDK12_ and _KDM6A_ occurred more frequently in the _ERG_-negative group. Recurrent gene fusions and

somatic mutations could explain only a subset of _ERG_-negative tumors, noting that more of these genes harbor somatic copy number alterations. Some of them are shared between the two

groups of tumors, others occurred more frequently in one group over the other. In addition to confirm several previous studies, we found novel recurrent SCNA for _ERG_-negative prostate

cancers, such as _ZNF292_ deletion. In summary, the _ERG_-negative group was found more heterogeneous in our study. When validated, the recurrently altered genes in specific patient groups

may contribute to better tumor stratification and prognosis. Among these genes, _MYC_ is a well-known oncogene that plays an important role in tumor progression. The amplification of _MYC_

is frequently observed in numerous human cancers57. In this study, we found that _MYC_ amplification frequency was significantly higher in the _ERG_-negative group. As expected, patients

with tumors harboring _MYC_ amplification show a strong association with poor outcome. Previous studies have reported that intact _CHD1_ is required for _ERG_ rearrangements in the process

of tumor initiation and deletion of _CHD1_ is mutually exclusive with ETS fusions58, that was consistently observed in our study. In addition to confirming known gene defects, we also

identified several novel prostate cancer associated genes which may play important roles in the tumorigenesis of _ERG_-negative cancer type. Our study highlights potentially actionable genes

which may provide opportunities for target therapy of _ERG_-negative prostate tumors. These findings include the frequent deletion of the tumor suppressor gene _FRK_ (6q22.1), a

tyrosine-protein kinase that negatively regulates cell proliferation59, in _ERG_-negative group (22.3% vs. 8.0%). Decreased expression of _FRK_ gene strongly correlated with its deletion.

Moreover, FRK protein could interact with known drugs and may have potential application in clinical practice60. Other potentially druggable genes including _WISP3_ (6q21), _LRP1B_ (2q22.1),

and _PRDM1_ (6q21)60,61. In total, 21 (34.4%) genes in our candidate gene list have potential clinical relevance, covering 66.7% of _ERG_ negative tumors. Interestingly, we found that

different gene alterations may result in similar expression change or pathway alteration. NOTCH signaling pathway is a typical example. Similar phenomenon has been observed in other cancer

types. Taken Wnt signaling pathway as an example, _TP53_, _CTNNB1_ and _AXIN1_ are important elements in Wnt signaling network; _CTNNB1_ is more frequently mutated in HCV-infected

hepatocellular carcinoma (HCC)62, while the mutations of _TP53_ and _AXIN1_ are more frequent in HBV-infected HCC63,64, which indicated different viral etiologies might activate Wnt

signaling in distinct ways. Increasing number of studies reports race/ethnicity differences in cancer research. Due to the lack of large-scale omics study of African and Asian prostate

cancer patients, directly comparisons among multiple races are challenging. Our focuse on the _ERG_-negative group could provide a reference for populations with low frequency of _ERG_

positive tumor types. Nine representative genes were sufficient to classify into sub-categories 67.7% _ERG_-negative tumors that was consistently seen in 77.5% of prostate cancers of African

American men. Our previous studies found that approximately 20% Chinese patients harbor _ERG_-positive tumors41. Therefore we are particularly interested in the frequently altered and

targetable genes in the _ERG_-negative tumor type. The validation of the genomic alteration and expression of these genes in Chinese patients is warranted. Accumulating data on _ERG_

negative prostate cancer will help to discover more disease progression associated and actionable driver genes. Additionally, further experimental assessments of the functional significance

for recurrent genomic and gene expression alterations are also warranted. Our study highlights new aspects of _ERG_-positive and _ERG_-negative prostate cancers at genomic, epigenetic, and

expression levels. In this study, multi-omics data integration provided a methodological reference to prioritize candidate CNV genes and to evaluate the effects of overall alterations. The

observed molecular differences on gene fusions, somatic mutations and copy number alterations between ERG-positive and ERG-negative prostate tumors suggest both common and distinct

mechanisms of prostate tumorigenesis. Genes with recurrent alteration may act as potential drivers and contribute to patient stratification into distinct prognostic or therapeutic groups.

These results will help experimental biologist and clinical doctors for further assessment of the functional significance of candidate genes. Together, our results provide new insights into

prostate tumorigenesis further refining the sub-classes of _ERG_-negative and _ERG_-positive prostate tumor types. METHODS DATA COLLECTION Somatic mutation (496 tumor samples), SCNA (492

tumor samples), methylation (497 tumor + 35 normal samples), and expression (497 tumor + 52 normal samples) data from TCGA primary prostate cancer cohort were used in this study65.

Clinically actionable genes and the interactions between genes and drugs were retrieved from DGIdb (http://dgidb.org/)60. PATIENT GROUP AND ETHNIC INFORMATION Samples were stratified into

ERG-positive and ERG-negative groups based on the combined ERG fusion evidences from TCGA research article (333 samples) and TFGDP database (http://www.tumorfusions.org/, 502 samples)5,46. A

patient was assigned to ERG-positive group if its ERG fusion was detected in either study. For genome wide fusion analysis and statistics except for ERG fusion, data from TFGDP database was

used. The ethnic information was collected from literature in which G. Petrovics _et al_. determine the ancestry of TCGA cohort by principal component analysis based on SNP genotype data49.

DETECTION OF SIGNIFICANTLY MUTATED GENES AND COPY NUMBER ALTERATIONS We used MutSigCV (version: 1.2) to detect significantly mutated genes for ERG-positive and ERG-negative groups,

respectively50. Chi-squared test and Fisher exact test (determined by theoretical frequencies and sample size) were used to test the significance of different alteration frequency between

the two groups. GISTIC 2.0 (version 6.10) was used to identify genomic regions that are significantly amplified or deleted in ERG-positive and ERG-negative groups, respectively51,66. To find

the common and specifically altered regions in the two groups, we divided the whole genome into consecutive bins (window length = 10 kb). For each bin, the SCNA status is determined by the

SCNA status of majority of bases in it (that is, longer than 5 kb). For arm-level SCNA regions, the frequency was estimated by the median frequencies of all bins in that region. Since the

significant SCNA regions usually contained huge genes, we focused on the copy number alterations of tumor suppressor genes (TSGs) and oncogenes. We obtained 1217 TSGs and 232 oncogenes from

TSGene Database (v2.0) and UniProtKB database (keyword:“Proto-oncogene [KW-0656]”)67,68. These genes were classified into two types based on the following filtering rules: 1) Common SCNA

genes: high frequency (>20%) in both ERG-positive and ERG-negative groups; 2) Group-specific SCNA genes: TSGs (Oncogenes) whose deletion (amplification) frequencies were significantly

different between two groups (P < 0.001) and the frequency difference was larger than 10%. SELECTION OF REPRESENTATIVE GENES FOR ERG-NEGATIVE GROUP We used genes with recurrent SCNA

(frequency >15%) or mutation (frequency >10%) as candidate feature genes for ERG-negative group. We defined a group of genes with higher priority: genes whose alteration frequency were

significantly higher in ERG-negative group than that in ERG-positive group, genes which were targetable or had interaction with drugs, and genes whose copy number alteration was

significantly correlated with expression. To remove genes with similar alteration pattern, we calculated Pearson correlation coefficient between genes and did unsupervised hierarchical

clustering. For each cluster, we selected genes with the highest frequency or higher priority as the representative genes. At last, six CNV genes and three mutation genes were selected as

final representative feature genes for ERG-negative group. OncoPrint was used to display the mutation landscape in ERG-negative group69. INDEPENDENT VALIDATION DATASET We used an independent

whole genome sequencing data (CPDR dataset) to validate the CNV candidate genes. The CPDR dataset including 7 ERG-positive and 7 ERG-negative prostate tumors. The Genomatix software

suite/NGS Analysis (http://www.genomatix.de) was used for CNV calling. DIFFERENTIAL EXPRESSION ANALYSIS We identified the differentially expressed genes among ERG-positive (n = 201),

ERG-negative (n = 296) and normal samples (n = 52). Normalized read counts were used to detect differential expression genes with R package voom and limma70. Genes with P value < 0.05 and

the absolute value of fold change (FC) > 2 were considered as differentially expressed. DIFFERENTIAL METHYLATION ANALYSIS We identified the differentially methylated genes among

ERG-positive (n = 201), ERG-negative (n = 296) and normal samples (n = 35) based on TCGA methylation data. Firstly, we removed the probes on X/Y/M chromosome or NA. Secondly, we found

diff-methylated sites with t-test p < 0.01 and the absolute difference of beta value > 0.2. Thirdly, we selected diff-methylated sites on promoter region (TSS200, TSS1500, 5’UTR and

1stExon). Fourthly, we retained methylation sites negatively correlated with the corresponding gene expression in a cis-regulatory manner. Fifthly, we concentrated on hyper-methylated sites

whose corresponding genes have significantly lower expression in tumor samples compared to normal samples. For the comparison between ERG-positive and ERG-negative group, genes

hyper-methylated in either group were taken into account. DATA AVAILABILITY All data generated or analyzed during this study are included in this published article (and its Supplementary

Information files). REFERENCES * Torre, L. A. _et al_. Global cancer statistics, 2012. _CA: a cancer journal for clinicians_ 65, 87–108 (2015). Google Scholar  * Barbieri, C. E. _et al_.

Exome sequencing identifies recurrent SPOP, FOXA1 and MED12 mutations in prostate cancer. _Nature genetics_ 44, 685–689 (2012). Article  PubMed  PubMed Central  CAS  Google Scholar  *

Barbieri, C. E. _et al_. The mutational landscape of prostate cancer. _European urology_ 64, 567–576 (2013). Article  PubMed  PubMed Central  CAS  Google Scholar  * Taylor, B. S. _et al_.

Integrative genomic profiling of human prostate cancer. _Cancer cell_ 18, 11–22 (2010). Article  PubMed  PubMed Central  CAS  Google Scholar  * Network, C. G. A. R. The molecular taxonomy of

primary prostate cancer. _Cell_ 163, 1011–1025 (2015). Article  CAS  Google Scholar  * Kan, Z. _et al_. Diverse somatic mutation patterns and pathway alterations in human cancers. _Nature_

466, 869–873 (2010). Article  ADS  PubMed  CAS  Google Scholar  * Berger, M. F. _et al_. The genomic complexity of primary human prostate cancer. _Nature_ 470, 214–220 (2011). Article  ADS 

PubMed  PubMed Central  CAS  Google Scholar  * Tomlins, S. A. _et al_. Recurrent fusion of TMPRSS2 and ETS transcription factor genes in prostate cancer. _Science_ 310, 644–648 (2005).

Article  ADS  PubMed  CAS  Google Scholar  * Tomlins, S. A. _et al_. TMPRSS2: ETV4 gene fusions define a third molecular subtype of prostate cancer. _Cancer research_ 66, 3396–3400 (2006).

Article  PubMed  CAS  Google Scholar  * Helgeson, B. E. _et al_. Characterization of TMPRSS2: ETV5 and SLC45A3: ETV5 gene fusions in prostate cancer. _Cancer research_ 68, 73–80 (2008).

Article  PubMed  CAS  Google Scholar  * Rubin, M. A., Maher, C. A. & Chinnaiyan, A. M. Common gene rearrangements in prostate cancer. _Journal of Clinical Oncology_ 29, 3659–3668 (2011).

Article  PubMed  PubMed Central  CAS  Google Scholar  * Robinson, D. _et al_. Integrative clinical genomics of advanced prostate cancer. _Cell_ 161, 1215–1228 (2015). Article  PubMed 

PubMed Central  CAS  Google Scholar  * Camacho, N. _et al_. Appraising the relevance of DNA copy number loss and gain in prostate cancer using whole genome DNA sequence data. _PLoS genetics_

13, e1007001, https://doi.org/10.1371/journal.pgen.1007001 (2017). Article  PubMed  PubMed Central  CAS  Google Scholar  * Attard, G. _et al_. Prostate cancer. _The Lancet_ 387, 70–82,

https://doi.org/10.1016/S0140-6736(14)61947-4 (2016). Article  Google Scholar  * Attard, G. _et al_. Prostate cancer. _Lancet (London, England)_ 387, 70–82,

https://doi.org/10.1016/s0140-6736(14)61947-4 (2016). Article  Google Scholar  * Hermans, K. G. _et al_. TMPRSS2: ERG fusion by translocation or interstitial deletion is highly relevant in

androgen-dependent prostate cancer, but is bypassed in late-stage androgen receptor–negative prostate cancer. _Cancer research_ 66, 10658–10663 (2006). Article  PubMed  CAS  Google Scholar 

* Lin, B. _et al_. Prostate-localized and androgen-regulated expression of the membrane-bound serine protease TMPRSS2. _Cancer research_ 59, 4180–4184 (1999). PubMed  CAS  Google Scholar  *

Tomlins, S. A. _et al_. Role of the TMPRSS2-ERG gene fusion in prostate cancer. _Neoplasia_ 10, 177IN171–188IN179 (2008). Article  CAS  Google Scholar  * Yu, J. _et al_. An integrated

network of androgen receptor, polycomb, and TMPRSS2-ERG gene fusions in prostate cancer progression. _Cancer cell_ 17, 443–454 (2010). Article  PubMed  PubMed Central  CAS  Google Scholar  *

Brase, J. C. _et al_. TMPRSS2-ERG-specific transcriptional modulation is associated with prostate cancer biomarkers and TGF-β signaling. _BMC cancer_ 11, 507 (2011). Article  PubMed  PubMed

Central  CAS  Google Scholar  * Ratz, L. _et al_. TMPRSS2: ERG gene fusion variants induce TGF-β signaling and epithelial to mesenchymal transition in human prostate cancer cells.

_Oncotarget_ 8, 25115 (2017). Article  PubMed  PubMed Central  Google Scholar  * Sreenath, T. L. _et al_. ETS related gene mediated androgen receptor aggregation and endoplasmic reticulum

stress in prostate cancer development. _Scientific reports_ 7, 1109 (2017). Article  ADS  PubMed  PubMed Central  CAS  Google Scholar  * Perner, S. _et al_. TMPRSS2-ERG fusion prostate

cancer: an early molecular event associated with invasion. _The American journal of surgical pathology_ 31, 882–888 (2007). Article  PubMed  Google Scholar  * Tandefelt, D. G., Boormans, J.,

Hermans, K. & Trapman, J. ETS fusion genes in prostate cancer. _Endocrine-related cancer_ 21, R143–R152 (2014). Article  CAS  Google Scholar  * Furusato, B. _et al_. ERG oncoprotein

expression in prostate cancer: clonal progression of ERG-positive tumor cells and potential for ERG-based stratification. _Prostate cancer and prostatic diseases_ 13, 228 (2010). Article 

PubMed  PubMed Central  CAS  Google Scholar  * Pettersson, A. _et al_. The TMPRSS2: ERG rearrangement, ERG expression, and prostate cancer outcomes: a cohort study and meta-analysis. _Cancer

Epidemiology and Prevention Biomarkers_ 21, 1497–1509 (2012). Article  Google Scholar  * Nam, R. K. _et al_. Expression of TMPRSS2: ERG gene fusion in prostate cancer cells is an important

prognostic factor for cancer progression. _Cancer biology & therapy_ 6, 40–45 (2007). Article  CAS  Google Scholar  * Demichelis, F. _et al_. TMPRSS2: ERG gene fusion associated with

lethal prostate cancer in a watchful waiting cohort. _Oncogene_ 26, 4596–4599 (2007). Article  PubMed  CAS  Google Scholar  * Perner, S. _et al_. TMPRSS2: ERG fusion-associated deletions

provide insight into the heterogeneity of prostate cancer. _Cancer research_ 66, 8337–8341 (2006). Article  PubMed  CAS  Google Scholar  * Fine, S. W. _et al_. TMPRSS2–ERG gene fusion is

associated with low Gleason scores and not with high-grade morphological features. _Modern pathology_ 23, 1325–1333 (2010). Article  PubMed  PubMed Central  Google Scholar  * Gopalan, A. _et

al_. TMPRSS2-ERG gene fusion is not associated with outcome in patients treated by prostatectomy. _Cancer research_ 69, 1400–1406 (2009). Article  PubMed  PubMed Central  CAS  Google

Scholar  * FitzGerald, L. M. _et al_. Association of TMPRSS2-ERG gene fusion with clinical characteristics and outcomes: results from a population-based study of prostate cancer. _BMC

cancer_ 8, 230 (2008). Article  PubMed  PubMed Central  CAS  Google Scholar  * Darnel, A. D., LaFargue, C. J., Vollmer, R. T., Corcos, J. & Bismar, T. A. TMPRSS2-ERG fusion is frequently

observed in Gleason pattern 3 prostate cancer in a Canadian cohort. _Cancer biology & therapy_ 8, 125–130 (2009). Article  CAS  Google Scholar  * Xu, B. _et al_. The prognostic role of

ERG immunopositivity in prostatic acinar adenocarcinoma: a study including 454 cases and review of the literature. _Human pathology_ 45, 488–497 (2014). Article  PubMed  CAS  Google Scholar

  * Petrovics, G. _et al_. Frequent overexpression of ETS-related gene-1 (ERG1) in prostate cancer transcriptome. _Oncogene_ 24, 3847 (2005). Article  PubMed  CAS  Google Scholar  * Hu, Y.

_et al_. Delineation of TMPRSS2-ERG splice variants in prostate cancer. _Clinical Cancer Research_ 14, 4719–4725 (2008). Article  PubMed  CAS  Google Scholar  * Cullen, J. _et al_.

Predicting Prostate Cancer Progression as a Function of ETS-related Gene Status, Race, and Obesity in a Longitudinal Patient Cohort. _European urology focus_ (2017). * Moniri, M. R., Hsing,

M., Rennie, P. S., Cherkasov, A. & Cox, M. E. The future of prostate cancer precision medicine: anti-ERG therapies. _Translational Cancer Research_ 6, S1136–S1138 (2017). Article  Google

Scholar  * Wang, X. _et al_. Development of Peptidomimetic Inhibitors of the ERG Gene Fusion Product in Prostate Cancer. _Cancer cell_ 31, 532–548. e537,

https://doi.org/10.1016/j.ccell.2017.05.001 (2017). Article  PubMed  PubMed Central  CAS  Google Scholar  * Mohamed, A. A. _et al_. Identification of a small molecule that selectively

inhibits ERG-positive cancer cell growth. _Cancer Res_, https://doi.org/10.1158/0008-5472.can-17-2949 (2018). * Sedarsky, J., Degon, M., Srivastava, S. & Dobi, A. Ethnicity and ERG

frequency in prostate cancer. _Nature Reviews Urology_. Preprint at https://doi.org/10.1038/nrurol.2017.140 (2017). * Magi‐Galluzzi, C. _et al_. TMPRSS2–ERG gene fusion prevalence and class

are significantly different in prostate cancer of caucasian, african‐american and japanese patients. _The Prostate_ 71, 489–497 (2011). Article  PubMed  CAS  Google Scholar  * Dobi, A. _et

al_. ERG-based stratification of prostate cancer highlights ethnicity associated biological differences. https://doi.org/10.1158/1538-7445.AM2015-5277 (AACR, 2015). * Ren, S. _et al_.

Whole-genome and transcriptome sequencing of prostate cancer identify new genetic alterations driving disease progression. _European urology_. Preprint at

https://doi.org/10.1016/j.eururo.2017.08.027 (2017). * Alumkal, J. J. & Herman, J. G. Distinct Epigenetic Mechanisms Distinguish TMPRSS2–ERG Fusion-Positive and-Negative Prostate

Cancers. _Cancer discovery_ 2, 979–981 (2012). Article  PubMed  PubMed Central  CAS  Google Scholar  * Hu, X. _et al_. TumorFusions: an integrative resource for cancer-associated transcript

fusions. _Nucleic Acids Research_. Preprint at https://doi.org/10.1093/nar/gkx1018 (2017). * Han, B. _et al_. A fluorescence _in situ_ hybridization screen for E26 transformation–specific

aberrations: identification of DDX5-ETV4 fusion protein in prostate cancer. _Cancer research_ 68, 7629–7637 (2008). Article  PubMed  PubMed Central  CAS  Google Scholar  * Esgueva, R. _et

al_. Prevalence of TMPRSS2–ERG and SLC45A3–ERG gene fusions in a large prostatectomy cohort. _Modern Pathology_ 23, 539–546 (2010). Article  PubMed  PubMed Central  CAS  Google Scholar  *

Petrovics, G. _et al_. A novel genomic alteration of LSAMP associates with aggressive prostate cancer in African American men. _EBioMedicine_ 2, 1957–1964 (2015). Article  PubMed  PubMed

Central  Google Scholar  * Lawrence, M. S. _et al_. Mutational heterogeneity in cancer and the search for new cancer-associated genes. _Nature_ 499, 214–218 (2013). Article  ADS  PubMed 

PubMed Central  CAS  Google Scholar  * Mermel, C. H. _et al_. GISTIC2. 0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human

cancers. _Genome biology_ 12, R41, https://doi.org/10.1186/gb-2011-12-4-r41 (2011). Article  PubMed  PubMed Central  Google Scholar  * Lee, J. H., Song, S. Y., Kim, M. S., Yoo, N. J. &

Lee, S. H. Frameshift mutations of a tumor suppressor gene ZNF292 in gastric and colorectal cancers with high microsatellite instability. _Apmis_ 124, 556–560 (2016). Article  PubMed  CAS 

Google Scholar  * Fabbri, G. & Dalla-Favera, R. The molecular pathogenesis of chronic lymphocytic leukaemia. _Nature Reviews Cancer_ 16, 145–162 (2016). Article  PubMed  CAS  Google

Scholar  * Liberzon, A. _et al_. The molecular signatures database hallmark gene set collection. _Cell systems_ 1, 417–425 (2015). Article  PubMed  PubMed Central  CAS  Google Scholar  *

Vogelstein, B. _et al_. Cancer genome landscapes. _Science_ 339, 1546–1558 (2013). Article  ADS  PubMed  PubMed Central  CAS  Google Scholar  * Myers, J. S., von Lersner, A. K., Robbins, C.

J. & Sang, Q.-X. A. Differentially expressed genes and signature pathways of human prostate cancer. _PloS one_ 10, e0145322, https://doi.org/10.1371/journal.pone.0145322 (2015). Article

  PubMed  PubMed Central  CAS  Google Scholar  * Beroukhim, R. _et al_. The landscape of somatic copy-number alteration across human cancers. _Nature_ 463, 899–905 (2010). Article  ADS 

PubMed  PubMed Central  CAS  Google Scholar  * Burkhardt, L. _et al_. CHD1 is a 5q21 tumor suppressor required for ERG rearrangement in prostate cancer. _Cancer research_ 73, 2795–2805

(2013). Article  PubMed  CAS  Google Scholar  * Hua, L. _et al_. FRK suppresses the proliferation of human glioma cells by inhibiting cyclin D1 nuclear accumulation. _Journal of

neuro-oncology_ 119, 49–58 (2014). Article  PubMed  CAS  Google Scholar  * Wagner, A. H. _et al_. DGIdb 2.0: mining clinically relevant drug–gene interactions. _Nucleic acids research_ 44,

D1036–D1044 (2015). Article  PubMed  PubMed Central  CAS  Google Scholar  * Cowin, P. A. _et al_. LRP1B deletion in high-grade serous ovarian cancers is associated with acquired chemotherapy

resistance to liposomal doxorubicin. _Cancer research_ 72, 4060–4073 (2012). Article  PubMed  CAS  Google Scholar  * Tornesello, M. L. _et al_. Mutations in TP53, CTNNB1 and PIK3CA genes in

hepatocellular carcinoma associated with hepatitis B and hepatitis C virus infections. _Genomics_ 102, 74–83 (2013). Article  PubMed  CAS  Google Scholar  * Schulze, K. _et al_. Exome

sequencing of hepatocellular carcinomas identifies new mutational signatures and potential therapeutic targets. _Nature genetics_ 47, 505–511 (2015). Article  ADS  PubMed  PubMed Central 

CAS  Google Scholar  * Totoki, Y. _et al_. Trans-ancestry mutational landscape of hepatocellular carcinoma genomes. _Nature genetics_ 46, 1267–1273 (2014). Article  PubMed  CAS  Google

Scholar  * Grossman, R. L. _et al_. Toward a shared vision for cancer genomic data. _New England Journal of Medicine_ 375, 1109–1112 (2016). Article  PubMed  Google Scholar  * Reich, M. _et

al_. GenePattern 2.0. _Nature genetics_ 38, 500–501 (2006). Article  PubMed  CAS  Google Scholar  * Zhao, M., Kim, P., Mitra, R., Zhao, J. & Zhao, Z. TSGene 2.0: an updated

literature-based knowledgebase for tumor suppressor genes. _Nucleic acids research_ 44, D1023–D1031 (2015). Article  PubMed  PubMed Central  CAS  Google Scholar  * Boutet, E. _et al_.

UniProtKB/Swiss-Prot, the manually annotated section of the UniProt KnowledgeBase: how to use the entry view. _Plant bioinformatics: methods and protocols_, 23–54,

https://doi.org/10.1007/978-1-4939-3167-5_2 (2016). * Gu, Z., Eils, R. & Schlesner, M. Complex heatmaps reveal patterns and correlations in multidimensional genomic data.

_Bioinformatics_ 32, 2847–2849 (2016). Article  PubMed  CAS  Google Scholar  * Law, C. W., Chen, Y., Shi, W. & Smyth, G. K. Voom: precision weights unlock linear model analysis tools for

RNA-seq read counts. _Genome biology_ 15, R29, https://doi.org/10.1186/gb-2014-15-2-r29 (2014). Article  PubMed  PubMed Central  CAS  Google Scholar  Download references ACKNOWLEDGEMENTS

This work was supported by the National Key Research and Development Program on Precision Medicine (2016YFC0902201, 2016YFC0902400), National Natural Science Foundation of China (31771472),

National Grand Program on Key Infectious Diseases (2015ZX10004801-005), and Chinese Academy of Sciences (ZDBS-SSW-DQC-02). AUTHOR INFORMATION AUTHORS AND AFFILIATIONS * CAS Key Laboratory of

Computational Biology, CAS-MPG Partner Institute for Computational Biology, Shanghai Institute of Nutrition and Health, Shanghai Institutes for Biological Sciences, University of Chinese

Academy of Sciences, Chinese Academy of Sciences, Shanghai, P. R. China Qingyu Xiao, Yidi Sun, Guo-Ping Zhao, Yixue Li & Hong Li * Center for Prostate Disease Research, Department of

Surgery, Uniformed Services University of the Health Sciences and Walter Reed National Military Medical Center, Bethesda, MD, USA Albert Dobi & Shiv Srivastava * Cancer Biomarkers

Research Group, Division of Cancer Prevention, National Cancer Institute, Bethesda, MD, USA Wendy Wang & Sudhir Srivastava * Department of Pathology, Zhongshan Hospital, Fudan

University, Shanghai, China Yuan Ji & Jun Hou Authors * Qingyu Xiao View author publications You can also search for this author inPubMed Google Scholar * Yidi Sun View author

publications You can also search for this author inPubMed Google Scholar * Albert Dobi View author publications You can also search for this author inPubMed Google Scholar * Shiv Srivastava

View author publications You can also search for this author inPubMed Google Scholar * Wendy Wang View author publications You can also search for this author inPubMed Google Scholar *

Sudhir Srivastava View author publications You can also search for this author inPubMed Google Scholar * Yuan Ji View author publications You can also search for this author inPubMed Google

Scholar * Jun Hou View author publications You can also search for this author inPubMed Google Scholar * Guo-Ping Zhao View author publications You can also search for this author inPubMed 

Google Scholar * Yixue Li View author publications You can also search for this author inPubMed Google Scholar * Hong Li View author publications You can also search for this author inPubMed

 Google Scholar CONTRIBUTIONS Q.X. and H.L. participated in the design of the study and drafted the manuscript; Q.X. and Y.S. collected the data and carried out data analysis; H.L. and Y.L.

conceived and directed the study; Y.J., J.H., A.D., S.S., W.W. and S.S. participated in the design and coordination of the study; H.L., Y.L. and G.Z. supervised the project. All of the

authors read and approved the final manuscript. CORRESPONDING AUTHORS Correspondence to Yixue Li or Hong Li. ETHICS DECLARATIONS COMPETING INTERESTS The authors declare no competing

interests. ADDITIONAL INFORMATION PUBLISHER'S NOTE: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ELECTRONIC

SUPPLEMENTARY MATERIAL SUPPLEMENTARY INFORMATION SUPPLEMENTARY TABLES 1-4-5 RIGHTS AND PERMISSIONS OPEN ACCESS This article is licensed under a Creative Commons Attribution 4.0 International

License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source,

provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons

license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by

statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit

http://creativecommons.org/licenses/by/4.0/. Reprints and permissions ABOUT THIS ARTICLE CITE THIS ARTICLE Xiao, Q., Sun, Y., Dobi, A. _et al._ Systematic analysis reveals molecular

characteristics of ERG-negative prostate cancer. _Sci Rep_ 8, 12868 (2018). https://doi.org/10.1038/s41598-018-30325-9 Download citation * Received: 04 March 2018 * Accepted: 27 July 2018 *

Published: 27 August 2018 * DOI: https://doi.org/10.1038/s41598-018-30325-9 SHARE THIS ARTICLE Anyone you share the following link with will be able to read this content: Get shareable link

Sorry, a shareable link is not currently available for this article. Copy to clipboard Provided by the Springer Nature SharedIt content-sharing initiative